CVE-2026-27893: vLLM's hardcoded trust_remote_code=True in NemotronVL and KimiK25 bypasses user security opt-out
Vulnerability summary
CVE-2026-27893 is a high (CVSS 8.8) Protection Mechanism Failure vulnerability in vllm-project vllm. EPSS estimates a 1.81% chance of exploitation in the next 30 days, higher than 78.4% of scored vulnerabilities.
Advisory lineage
Upstream 0 · Downstream 0 · Related 1Timeline
- 26-03-2026PublishedVulnerability first disclosed.
- 27-07-2026Last modifiedVulnerability information updated.
Description
vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.18.0, two model implementation files hardcode `trust_remote_code=True` when loading sub-components, bypassing the user's explicit `--trust-remote-code=False` security opt-out. This enables remote code execution via malicious model repositories even when the user has explicitly disabled remote code trust. Version 0.18.0 patches the issue.
CVSS metrics
4 scores · 4 sources · 1 version- OSV · PypiEcosystem advisory · calculated8.8HighCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
- OSV · ChainguardEcosystem advisory · calculated8.8HighCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
- NVDNIST8.8HighCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
EPSS trend
Probability
1.81%
Chance of exploitation in the next 30 days
Percentile
78%
Versus all other CVEs
30-day change
Increase +0.46 pp
From 1.35% on 22-09-2026
Known exploited
Compare KEV catalogsNot in the CISA KEV catalog.
Not in the ENISA CNW catalog.
Not in the CIRCL catalog.
Public exploits
0 foundNo public exploits indexed.
Weaknesses
CWE · root cause- CWE-693Protection Mechanism FailureThe product does not use or incorrectly uses a protection mechanism that provides sufficient defense against directed attacks against the product.
- CWE-501Trust Boundary ViolationThe product mixes trusted and untrusted data in the same data structure or structured message.
Affected systems
Chainguard5 products
- py3.10-vllm-cuda-12.4< 0.18.1-r0
- py3.12-vllm-cuda-12.4< 0.18.1-r0
- tritonserver-backend-vllm-cuda-12.9All versions (No fix recorded)
- tritonserver-backend-vllm-cuda-13.0< 25.11-r7
- vllm-openai-cuda-12.9< 0.19.0-r0
PyPI1 product
- vllm≥ 0.10.1, < 0.18.0
Vllm1 product
- vllm≥ 0.10.1, < 0.18.0
Vllm-project1 product
- vllm≥ 0.10.1, < 0.18.0
4 vendors · 8 products
References
21 links- github.com/vllm-project/vllm/security/advisories/GHSA-7972-pg2x-xr59X Refsource CONFIRMWEBAdvisoryVendor Advisory
- github.com/vllm-project/vllm/commit/00bd08edeee5dd4d4c13277c0114a464011acf72X Refsource MISCWEBFIXPatch
Show all 21 references
- github.com/vllm-project/vllmPACKAGE